Study On Impact Damage To Typical Fruits Based On Hyperspectral Imaging Technology

Fruits are rich in vitamins,minerals and dietary fiber,which have always been favored by people and have become an important part of the human diet.However,fresh fruits are very susceptible to mechanical damage during harvesting,packaging and transportation,which can result in a significant reduction in fruit quality.Among various forms of damage,impact damage is the most severe and most likely to occur.If damage occurs,the economic loss can be reduced by classifying the fruit based on the degree of the damage.Thus,an objective and quantitative assessment of the damage's degree is required.However,this still remains an important challenge in the field of food safety.The traditional artificial sensory detection and destructive prediction evaluation methods are time-consuming and laborious,and the precision is not high,which cannot meet the needs of rapid,non-destructive and real-time automatic detection and classification of modern agriculture.In this paper,the typical fruits including mango and apple were taken as the research object.Based on hyperspectral imaging technology,combined with a variety of spectral analysis techniques and mathematical modeling methods,the present study observed the changes of spectra,the physical and chemical attributes and corresponding mechanical parameters in terms of fruits subjected to impact damage.Then the main contributing factors affecting the degree of damage were found and the mathematical models between fruits'spectra information and the physical and chemical attributes as well as mechanical parameters were further established.The overall study achieved the non-destructive identification on whether the fruits are damaged and the non-destructive prediction of fruits'quality and other parameters.It provides an important reference for non-destructive evaluation of mechanical damage to fruits.The main research contents and results are as follows:(1)Near-infrared hyperspectral imaging technology(900-1700nm)was used to predict the physical and chemical attributes of mango as well as the non-destructive identification of damage degree.Hyperspectral images of mango samples dropped from different heights(0.5 m,1.0 m,1.5 m)were collected and the average spectra of the region of interest(ROI)of the samples was extracted.The characteristic wavelengths of the whole spectra were selected as the spectral variable in combination with a competitive adaptive reweighting algorithm(CARS).Physical and chemical attributes of all samples were measured by physical and chemical methods.The partial least squares regression(PLSR)model between samples'spectral variables and their corresponding physical and chemical attributes was established.For pulp firmness(PF),total soluble solids(TSS),titratable acid(TA)and color(?b*),the coefficients of determination~2 and the corresponding mean square error RMSEP from PLSR model were 0.84 and 3.16 N,0.9 and 0.49 ~oBrix,0.86 and 0.07%,0.94and 0.96,respectively.Finally,according to the maturity index RPI and discriminant analysis(DA),the degree of impact damage of the sample is classified and the accuracy of the classification result is not less than 77.8%.(2)Near-infrared hyperspectral imaging technology(900-1700nm)was used to quantify and predict the impact damage area and other mechanical parameters of the apple samples during impact process.Hyperspectral images of apple samples dropped from different heights(0.5 m,1.0 m,1.5 m)were collected and the average spectra of the region of interest(ROI)of the samples were extracted.The characteristic wavelengths of the whole spectra were selected in conjunction with the regression coefficient(RC)method.The damage area(DA),contact load(CL)and absorbed energy(AE)of the apples during the falling process were obtained by means of pressure sensitive film technology and a high speed camera.The pulp firmness(PF)of the sample was extracted from the puncture test of the apple.The statistical analysis shows that there is a significant linear correlation between the samples'absorbed energy,the contact load and the corresponding damage area.The determination coefficients~2were 0.93 and 0.92,respectively,which indicates that the mechanical parameters can well characterize the impact damage of apple.Finally,a partial least squares regression(PLSR)model between spectral variables and the damage areas as well as mechanical parameters was established.For the damage area,absorbed energy,contact load and pulp firmness,the coefficient~2and the root mean square error RMSEP measured by the PLSR prediction model reached 0.8 and 116.73 mm~2,0.89 and 0.075 J,0.53 and 67.38 N,0.65 and 19.99 g,respectively.